Where essential functionality can be trivially reproduced and customized to specific lawyers' needs or tastes, it seems to me that the way software engineers can be of most use is by offering to review the vibed code for security concerns, and to suggest refactoring for efficiency (maintenance now being less of a concern). Still, most lawyers' use of vibed software will not be on so grand a scale that efficiencies would be much of a concern either (large discovery reviews or legal-corpus processing as exceptions, perhaps).
Furthermore, it seems like a disincentive to create and offer software to the broad legal market: even a description of the idea of the functionality, let alone a video or trial version, would be enough to copy and tweak the software. We may end up with a Cold War and Balkanization of software: FOSS on the one hand, and much non-FOSS in the form of "homemade" (vibed), unshared, unsold custom software siloed in each firm.
As everyone learns to eke out decent results from LLMs, which should become commodities themselves, the real differentiators for law firms will be creativity in designing workflows and, as always, human legal acumen.
